I'm attempting to set up a test HDD for the company I work for. I need to set up 150+ bootable partitions for our Mobo final test technicians. With a week of failed attempts using WinXP, I'm looking into Linux options. Someone has recommended this route to me and to use Grub as my boot loader. Anyone out there have thoughts or suggestions concerning this task. And more importantly, has anyone ever seen the limitations of Grub or other boot loaders fully tested or exceeded? A end result user friendly boot menu with 150+ Mobo models is what I'm looking to accomplish. All suggestions welcome. Thanks.

I'm not aware of a limitation in grub but with 150 entries it could take some time to navigate.

A quick way to ease the picking of a boot option may be to boot a minimal kernel with a filesystem on an initrd which just presents an option to pick a manufacturer and then an exact mobo version. This then sets that boot option as the default and restarts the machine.

Thanks for the suggestion. All our models here are listed by part numbers so I'd hope I could just sort the menu alphabetically or something. However I'd love to further explore what you mentioned. Any chance you could point me towards some other threads that talk about that in more detail? My only linux experience as of yet was installing red-hat a few years ago in one of my classes at school.

Yes, you could obviously sort the menu and it may be enough for your needs.

I'm not aware of any threads suggesting exactly what I mentioned but you would probably want to consider using busybox as the basis for your initrd as it has all of the basic tools required.

I have over 120 partitions in the box and Grub is booting 50+ systems

I am aware of Linux permits only 255 raw devices
An IDE can have 63 partitions but I can only manage 60
A Sata/SCSI may not have more than 15 partitions.

XP can boot a maximum of 10 when I tried.

Lilo can do 15 because of its static menu space like NTLDR

My guess for Grub is no limit because Grub unattached to a system can boot another Grub to form a chain. Thus one can have one Grub booting say 10 systems with a link to a 2nd Grub booting the next 10 systems and so on.

I am already giving one GRub to manage my M$ systems, one for Large Linux, one for recent additions and one for all systems together. These Grubs work in small partitions with only stage1, stage2 and menu.lst files inside.

Lastly as far as I am aware only 0.97 version of Grub (current version) can work beyond the 137Gb barrier and that will be needed if you use large hdds.

I've done a bit more research and perhaps hit a stumbling block today as I prepare for this test HDD. I keep reading something about a max number of partitions being 63?? Is this correct? Is there a way around it? I'll be dealing with very large GB HDD's, and am hoping to set up all 150+ partitions on one HDD. Anyone see possible problems or flaws in that thought process?

More reading makes me think its 63 partitions per every extended partition that I create. That seem right to anyone?

I wrote in my last message

"An IDE can have 63 partitions but I can only manage 60"

Linux partitioning program cfdisk simply stops creating any partition after 60.

Also don't fool yourself in thinking more than one extended partition is possible in a hdd when working to the PC standard.

BIOS only reads 16 bytes times 4 after the 447th byte of the MBR for the 4 primary partitions. If anybody creates an extended partition one primary partition must be given up.

The extended partition has no usuable space inside is just a border for the logical partitions.

Logical partitions are a continuous chain in which the ith partition carries the address of the i+1th partition. Physically you can't have more than two consecutive chains in a hard disk. You will notice too that the partitioning program will automatically grab all the area to form the extended partition once the first primaries have been created.

If Linux does not read beyond the 60th (or 63th) partition you can kiss good bye your idea of 150-partition in one disk and concentrating the job with 3 IDE disks.

It is possible that Linux permits 63 instead of 60 partitions. The disk I tried to get 60+ partition did have a FreeBSD and a Solaris in it. The Unix operating systems use a "slice" system to further subdivide the primary partitions and this feature may have triggered Linux to believe that I hit the ceiling limit of the quoted 63 partitions.

I missed the part about the number of partitions, I've got to ask though - why would you need 150 partitions? Why not just have 150 kernels and use the same partition?

~David~

Can you explain 150 kernels better? That threw me for a loop. This is my first experience with Linux so I'm newbie extreme here.

My concern for the 150 partitions is this. Drivers specific to the test procedures of each board must remain untouched and stable from other board models. The problem came up before that several boards on one partition was causing instability and some drivers were uninstalling previous ones etc. etc. So there is only a need to keep each instal 100% separate. If there is another way other than individual partitions its something I need to explore. I don't want to be on my 60'th or so partition and then Linux prevent me from creating new ones.

Can other partitioning tools then create additional needed partitions for me to install on? My problem isn't not understanding the actual partitioning but what Linux can handle and deal with. I'm all about trial and error, but I can't waste a few months getting up to 60's and then tell my bosses that I've hit another wall and the project has to be scrapped and re-engineered again.

I appreciate ideas and help from all....I realize this is an odd project and there really aren't examples of it being done before but if anyone knows obstacles or issues that may come up and advice they could give on how to prepare for those things I'd be grateful.

Thanks again all....

Ksheep

There is nothing to stop you having X number of kernels all using the same root filesystem so long as you give them all different localversion strings (probably the model of the motherboard). This way you could just set up a generic system and load whatever modules you want based on the kernel you boot.

Look it takes about 10 minutes to set up 60 partitions and also to find out if you can cross the 63 barrier.

If you are doing it repeatedly you should be able to write a script file to automate the process for formatting and loading up a system to each partition, assuming you load one standard system for all the partition.

David's idea of loading different kernel is also a good choice but you may need to load a large number of systems to get different kernels.